Every model, one place

OpenAI's fastest, cheapest GPT-5.6 tier — set as Bonxai's default. Strong for high-volume, well-defined tasks; Terra/Sol are the more capable (pricier) tiers in the same family.

Automatically routed to whichever free model OpenRouter has healthy right now — the most drift-resistant default, since it isn't pinned to one vendor's slug.

NVIDIA's largest open-weight model — strongest free option for hard questions.

Mid-size NVIDIA open-weight model, a faster tradeoff against Ultra.

Small, reasoning-tuned NVIDIA model — shows its work on harder problems.

Google's current-generation open-weight model, general-purpose.

Smaller mixture-of-experts Gemma 4 variant — quicker responses.

Small, fast open-weight model — good for quick, low-latency answers.

General-purpose model from Thinking Machines Lab.

Code-leaning model from Poolside — a solid pick for programming questions.

Cohere's small code-focused model.

Available via OpenRouter, but not free — needs a funded OpenRouter account balance before it'll work. Marked coming-soon until that's set up.

Anthropic's flagship model. Rolling out on a direct enterprise key.

OpenAI's flagship reasoning model. Pending enterprise access.

Smaller, cheaper GPT-5 tier. Pending enterprise access.

Google's flagship Gemini model. Pending enterprise access.

Fast, low-cost Gemini tier. Pending enterprise access.

Google's fast image-generation/editing model. Pending enterprise access.

xAI's flagship model. Pending enterprise access.

Search-grounded answers with citations. Pending enterprise access.